Re: What to do with archived redo-logs ?

Graham Miller wrote:
>
> "John P. Higgins" <jh33378nospam_at_deere.com> wrote (in
> <34B6B6C2.60BD_at_deere.com>)...
>
> |
> | Maybe I said this too strongly. If you do a hot backup after restarting
> | archivelog mode and the system crashes before the backup is complete,
> | then the only backup you have is prior to the big batch job. You would
> | have to restore and recover up to the last available log and rerun your
> | batch job. If you can accept this (admittedly small) risk, go with your
> | normal hot backup.
> |
> | If the hot backup completes ok, then it is certainly sufficient.
> Hello,
> I am a little confused about the 'hot backup' discussed in this
> thread. I am not a DBA.
> My understanding is that when running in 'Archive Log' mode then redo
> logs are copied to a separate directory. In 'non-archive log' mode
> then Oracle overwrites the redo-logs.
> Now, if that is the case, then if the batch job causes redo-logs to
> be overwritten, then how does a successful 'hot backup' after
> restarting the database in 'Archive log' mode allow a complete
> recovery from a time before the batch job started, since some redo
> transactions will be missing?
> Have i missed the point somewhere?
>
> graham

No, that IS the point. The batch job is un-recoverable because it executed in NOARCHIVELOG mode. in fact, everything from the start of NOARCHIVELOG mode until the successful backup (hot or cold) is un-recoverable.

So, if your disk crashed while doing the hot backup, the only backup available is from before the batch job. You would need to restore from that backup and recover up to the end of the archived logs. Then you would be back to the same point where you could re-run the batch job. You would then restart in ARCHIVELOG mode and start a backup. Any work done during the failed backup would be lost.

What is the risk of a disk crash during the hot backup in your shop? Probably pretty small. Can you live with that risk? We accept the risk on our development systems, but not on our production systems: we use a cold backup. Received on Sat Jan 10 1998 - 00:00:00 CST